Affect in Human-Robot Interaction

More and more, robots are expected to interact with humans in a social, easily understandable manner, which presupposes effective use of robot affect. This chapter provides a brief overview of research advances into this important aspect of human-robot interaction. Keywords: human-robot interaction, affective robotics, robot behavior I. Introduction and Motivation Humans possess an amazing capability of attributing life and affect to inanimate objects (Reeves and Nass 96, Melson et al 09). Robots take this to the next level, even beyond that of virtual characters due to their embodiment and situatedness. They offer the opportunity for people to bond with them by maintaining a physical presence in their world, in some ways comparable to other beings, such as fellow humans and pets. This raises a broad range of questions in terms of the role of affect in human-robot interaction (HRI), which will be discussed in this article: • What is the role of affect for a robot and in what ways can it add value and risk to humanrobot relationships? Can robots be companions, friends, even intimates to people? • Is it necessary for a robot to actually experience emotion in order to convey its internal state to a person? Is emotion important in enhancing HRI and if so when and where? • What approaches, theories, representations, and experimental methods inform affective HRI research? I.1 Roles of emotion in robotics There exist at least two different roles for emotion in robotic systems. The first, which will only be briefly discussed, is to serve as an adaptive function that increases the probability of correct behavior, some of which may relate to survival of an agent (human or robotic) in its environment. The second is for the benefit of the human when interacting with a robot by providing a means and mechanism for increasing the bandwidth in communication using nonverbal methods to create a more effective and stronger relationship between artifact and person. I.1.1 Adaptive behavior for survival Moravec (1988) notes that humans may even perceive emotions in robots even without deliberately modeling them: for example, if a robot backs away from a staircase it might be interpreted as a fear of falling by a person observing. Braitenberg (1984), using a series of Gedanken (thought) experiments, also demonstrates that vehicles can exhibit love, fear, and aggression that is attributed to them solely by virtue of human observation. These perceived emotions are likely to attune the robot more closely to its environment and thus enhance its survivability, but are not geared expressly for human-robot interaction. People have a natural propensity to anthropomorphize artifacts (Reeves and Nass 1996) even if there was no deliberate intent by the designer to do so. I.1.2 Human-­‐robot interaction Many other researchers, some of which are discussed in more detail below, have chosen to deliberately embed explicit models of affect into robots, with the express purpose of enhancing the relationship between the human and robot, and in some cases with the explicit goal of fostering a strong attachment by a person to the artifact. The underlying goal here is to produce a robotic platform that can be a friend or even a life-long companion to a human (Arkin et al 2003), and in some cases even approach the possibility of intimate human-robot relations (Levy 2008). I.2 Definitions in Context Definitions for nebulous affective terms such as emotions can be debated ad infinitum (Arkin 05). This volume undoubtedly addresses this in other chapters. We should note, however, we take a solipsist stance, i.e., that robots do not need to experience affective phenomena in the same way as humans do or even at all, in order for them to be perceived as possessing them. So no claim is made that the robot actually experiences emotions, but rather that the goal of affective human-robot interaction is to convey the perception to a person that it does. While this may be unsatisfying to a philosopher, it is a pragmatic solution to the roboticist, where affect lies in the eye of the beholder. I.3 A few short exemplars In order to carry out this illusion, many psychological models of human affect have been explored. Two examples that have had commercial success are described.